Robust estimator based on information potential and its application to simultaneous localization and mapping

We present a novel robust estimator based on information potential optimization techniques and apply it to simultaneous localization and mapping on segment-based maps. Structured indoor environment can be efficiently described with Segment-based maps. Usually, in dynamic environment, sample data collected by range-finders suffer from noises and disturbances. Sample data are divided into clusters with split-and-merge. Inliers of the segment are selected according to the information contribution which is measured by information potential. After the local map is built, particle filters are adopted to update robot poses and maps. The recursive information potential reduces computations of information contribution of each sample. Simulations and experimental results validate the strong robustness of the proposed estimator, and the accuracy and efficiency of the proposed strategy based on the robust estimator.